We Are The New Farmers 2024 Automation · Content Systems

Research-to-Blog Post Assistant: A Year of Content in One Afternoon

How I built an end-to-end content pipeline that turns a single research topic into a fully drafted, SEO-optimized, brand-aligned blog post in under 5 minutes.

60-90 min Previous time per post from research to draft
Under 5 min Time per post with the system
1 afternoon To produce a full year of content

The Problem

Customers buying functional food products increasingly want to understand the science behind what they're eating. For We Are The New Farmers, that meant a growing demand for high-quality, scientifically backed content about spirulina: its health benefits, its functional properties, and its role in specific conditions.

The problem was capacity. Writing a single well-researched post meant finding credible sources, synthesizing them into an outline, drafting in our brand voice, and optimizing for the right keywords. At 60 to 90 minutes per post, building a meaningful content library would take months of manual work. We needed a different approach.

"The constraint wasn't ideas or knowledge. It was the time cost of translating research into publish-ready content at any reasonable volume."

What I Built

An automated pipeline that takes a single topic as input and returns a fully drafted, Shopify-ready blog post as output. The system handles research aggregation, source management, SEO integration, and brand voice alignment without any manual steps between input and draft review.

The output is better than what I was producing manually. Posts are more thoroughly sourced, more consistently structured, and already formatted with HTML tags for direct paste into Shopify's blog interface. One afternoon of running the pipeline generated enough content to cover the rest of the year.

How It Works

The Pipeline, Step by Step

01
Input: One research topic

The only required input is a topic. I used ChatGPT to convert raw SEO keywords into richer, more specific research prompts that yield better downstream content.

Example
Keyword: "What is Spirulina"
Input: "Current research and historical context of spirulina: what it is, where it comes from, and how it's used in modern wellness"
02
Query Refiner: Expand one topic into five angles

The query refiner takes the input and generates five distinct search queries to ensure comprehensive coverage of the topic from multiple research angles.

Example output for "Spirulina and Thyroid Health"
spirulina effects on thyroid function spirulina supplementation and thyroid health spirulina impact on thyroid hormone levels studies on spirulina and thyroid outcomes spirulina benefits and risks for thyroid patients
03
Research Agent: Deep research via OpenRouter

The five queries are passed via HTTP request to OpenRouter, which routes them to the best available model for deep research. The agent returns a structured summary and a list of primary sources used.

04
Airtable: Research aggregation database

Every research request automatically creates a new record in Airtable storing the topic, the OpenRouter summary, and all citations. This doubles as a reusable knowledge base for future agents and content workflows.

05
Blog Writing Agent: Draft in brand voice

The writing agent converts the research summary into a New Farmers blog post. Its system prompt contains full content structure guidelines, HTML formatting rules for Shopify, brand voice samples, product integration instructions, scientific accuracy requirements, and SEO keyword logic. The agent autonomously selects the most relevant keywords and flags them in the output for human review.

06
Output: Google Doc ready for review and publish

The workflow creates a new Google Doc titled after the blog post, containing the full draft formatted for direct paste into Shopify. Human review remains the final step before publishing.

Tools and Stack

What I built it with

n8n
Workflow automation layer. Orchestrates the full pipeline from topic input through Google Doc creation.
OpenRouter
Unified interface for accessing multiple LLMs. Used for the deep research phase to route queries to the best available model.
Airtable
Research aggregation database. Stores every topic, summary, and citation set as a reusable knowledge base.
GPT-4o mini
Blog writing agent. Converts research summaries into structured, brand-aligned, SEO-optimized draft posts.
Google Docs
Final output format. Each post is auto-created as a named Doc, formatted for direct paste into Shopify.
Google Ads
Keyword research input. Used manually to identify high-value SEO terms that get fed into the writing agent's instructions.

Watch It in Action

A full walkthrough of the pipeline from topic input to finished Google Doc output.

Lessons Learned

01 The research database is the real asset. The Airtable aggregation layer turned out to be more valuable than the blog posts themselves. It became a knowledge base that other agents and workflows can draw from, making the system compound in value over time.
02 Instructions belong in the system prompt, not as tools. Passing the blog writing guidelines as a tool caused the agent to consistently ignore them. Moving everything directly into the system message solved it immediately. Less elegant architecturally, but it works reliably.
03 Source verification is still a human job. The deep research agent occasionally surfaced links pointing to competitor websites. Human review before publishing remains essential, particularly for scientific claims and external links.
04 Volume is only half the win. The posts produced were higher quality than previous manual ones, not just faster to create. Forcing a consistent structure and sourcing standard through the system prompt raised the floor across every piece of content.